Proximity-dependent and proximity-independent trans-splicing in mammalian cells

Abstract

Most human pre-mRNAs are cis-spliced, removing introns and joining flanking exons of the same RNA molecule. However, splicing of exons present on separate pre-mRNA molecules can also occur. This trans-splicing reaction can be exploited by pre-trans-splicing molecules (PTMs), which are incapable of cis-splicing. PTM-mediated trans-splicing has been utilized to repair mutant RNAs as a novel approach to gene therapy. Herein we explore how the site of PTM expression influences trans-splicing activity. We stably inserted a PTM expression cassette into the genome of HEK293 cells, generating clonal lines with single, unique insertion sites. We analyzed trans-splicing to the gene where the PTM was integrated, as well as genes neighboring these loci. We observed some pre-mRNAs only serve as substrates for trans-splicing when they are expressed in immediate proximity to the PTM expression site. The need for PTMs to be in close proximity with pre-mRNAs to trans-splice with them is consistent with the observation that pre-mRNA cis-splicing occurs cotranscriptionally. Interestingly, we identified several cellular pre-mRNAs in one localized area that serve as trans-splicing substrates irrespective of the PTM expression site. Thus, we find multiple cellular pre-mRNAs require PTM expression in close proximity to trans-splice while others do not.

INTRODUCTION

A growing body of evidence supports the concept that transcription is coupled to cis-splicing of pre-mRNAs (Maniatis and Tasic 2002; Kornblihtt et al. 2004). Using affinity chromatography on HeLa cell extracts, with immobilized transcription elongation factor, splicing factors copurify with RNA polymerase II (Robert et al. 2002). This same group (Robert et al. 2002) showed the reverse is also true—this core RNA polymerase II and its general transcription factors coimmunoprecipitate with an anti-splice factor antibody. Wetterberg et al. (2001) were able to identify and reconstruct the three-dimensional (3D) structure of the active Balbiani ring 3 (BR3) in Chironomus tentans. Using electron microscopy, Wetterberg et al. (2001) showed both a splicing factor (U2 snRNP) and a hyperphosphorylated RNA polymerase II (pol II) C-terminal domain (CTD) were within the same complex along each growing BR3 pre-mRNA transcript. This observation supports earlier findings showing areas of active transcription colocalized with both pol II and SR splicing proteins (Zeng et al. 1997) or the family of spliceosomal snRNPs (Rohdewohld et al. 1987). In vitro studies show a hyperphosphorylated pol II CTD activated splicing in a concentration-dependent manner (Hirose et al. 1999), and truncating the CTD inhibited splicing in vivo (McCracken et al. 1998). Cotranscriptional splicing is necessary in the case of very long genes, such as dystrophin (Tennyson et al. 1995). As shown in this work, the dystrophin transcript requires ∼16 h to synthesize; thus post-transcriptional splicing is not used for all introns and would not be practical. Coupling transcription and mRNA processing may also be advantageous; Roberts et al. (1998) has suggested concentrating the proteins involved in processing would increase efficiency of transcript production. Given the close proximity of the processes of transcription and cis-splicing, we wondered if proximity may also influence trans-splicing.

In eukaryotes, most genes are interrupted by introns. During processing of the pre-mRNA, these introns are removed and the remaining exons are joined. This splicing of exons takes place in the nucleus via two trans-esterification reactions catalyzed by the spliceosome. For spliceosome-mediated splicing to proceed, pre-mRNA substrates require a 5′ splice site (5′ss), branchpoint sequence (BPS), polypyrimidine tract (PyT), and a 3′ splice site (3′ss) (Hastings and Krainer 2001; Reed 2003; Tazi et al. 2005). In mammalian cells, this process of intron removal and exon ligation usually occurs such that exons from a single RNA molecule are joined together. In contrast to this common process of cis-splicing, rare cases of trans-splicing have been observed.

In trans-splicing, RNA exons from two separate pre-mRNA molecules are joined to create a single chimeric mRNA (Maniatis and Tasic 2002; Horiuchi and Aigaki 2006). In mammalian cells, the pre-mRNA molecules may be transcripts from the same gene (Caudevilla et al. 1998; Akopian et al. 1999; Takahara et al. 2000; Flouriot et al. 2002) or from different genes (Li et al. 1999; Finta and Zaphiropoulos 2002; Hirano and Noda 2004). One pre-mRNA molecule provides the donor 5′ss, and the second pre-mRNA molecule provides the BPS, PyT, and 3′ss. The trans exon ligation proceeds through the same trans-esterification steps carried out in cis-splicing, whereby the BPS of one transcript initiates the nucleophilic attack on the 5′ss of the other transcript. Though not typically utilized, mammalian cells are capable of trans-splicing. Over 20 yr ago was the first in vitro demonstration that two separate RNA molecules can be joined in a trans splicing process (Konarska et al. 1985). This reaction was first shown to take place in vivo in rat cells between separate viral pre-mRNA transcripts (Eul et al. 1995). Later, Caudevilla et al. (2001b) demonstrated trans-splicing can also occur between viral and cellular transcripts. Moreover, several hybrid, or trans-spliced, cellular mRNAs have been isolated from human tissues (Finta and Zaphiropoulos 2002). Trans-splicing has been speculated to exist as an additional method of increasing protein diversity (Caudevilla et al. 2001a; Finta and Zaphiropoulos 2002; Takahara et al. 2002, 2005), though in mammals, a functional protein created through trans-splicing has yet to be found (Horiuchi and Aigaki 2006). It is worth mentioning, however, the bursicon gene in mosquitoes (Robertson et al. 2007) and two Drosophila genes—mod(mdg4) and lola—undergo trans-splicing to produce essential proteins (Dorn and Krauss 2003; Horiuchi et al. 2003).

The cell's ability to trans-splice two pre-mRNA molecules led to a new field in RNA therapeutics. Puttaraju et al. (1999) was the first to design a targeted precursor trans-splicing RNA molecule (PTM) that could use the endogenous spliceosome machinery to modify a specific cellular RNA transcript. Since then several studies have used the PTM method to repair mutant RNA transcripts (Kikumori et al. 2001; Liu et al. 2002; Rogers et al. 2002; Chao et al. 2003; Lai et al. 2005; Rodriguez-Martin et al. 2005; Coady et al. 2007). In addition to correcting the intended mutant RNA, however, trans-splicing can also occur with unintended cellular RNA transcripts (Kikumori et al. 2001; Rusconi 2001; Crystal 2002).

RESULTS

In our studies, we examined how proximity would impact a 3′ pre-trans-splicing molecule designed to react with any 5′ donor site on a pre-mRNA it encountered. We constructed an expression cassette with a pre-trans-splicing molecule (pCIZ) that contains the splicing features necessary for optimal trans-splicing activity: an intron region that includes the branchpoint sequence (TACTAAC, underlined A is the reactive nucleotide), a PyT, a 3′ss at the start of an exon (a fragment of LacZ), and a polyadenylation signal from the rat pre-proinsulin gene (PPI) transcribed from a CMV (cytomegalovirus immediate early) promoter (Fig. 1A). The transcript from the trans-splicing molecule will use the endogenous spliceosome and trans-splice the LacZ/PPI region onto nearby cellular RNAs (Fig. 1B). These spliceosome-mediated trans-splicing reactions compete with cis-splicing (Fig. 1B). The trans-spliced RNAs can then be identified as they contain a LacZ/PPI “tag” sequence.

FIGURE 1.

(A) Trans-splicing expression cassette (pCIZ). Diagram of the untargeted trans-splicing expression cassette pCIZ. The positions of the CMV promoter, branchpoint sequence (BPS), polypyrimidine tract (PyT), 3′splice site (3′ss), 3′ exon (partial LacZ) tag, and the rat pre-proinsulin (PPI) polyadenylation signal are all noted. The blue “A” within the BPS begins the nucleophilic attack on the 5′ss to begin the splicing reaction. The intronic sequence is derived from adenovirus type II. (B) Trans-splicing competes with cis-splicing reactions. Trans-splicing of the LacZ tag onto cellular RNA transcripts takes place within the endogenous spliceosome and competes with cis-splicing. For both, splicing of exons and removal of introns proceeds via two trans-esterification reactions (steps 1 and 2 for cis-splicing and steps 1′ and 2′ for trans-splicing. A trans-spliced pre-mRNA transcript will be “tagged” with the LacZ as its 3′ exon. (C) Diagram of the retroviral construct with the 3′ PTM cloned into the U3 region of the LTR. The trans-splicing molecule (pCIZ) was PCR amplified with primers that included two restriction enzyme sites—BglII and MluI. These sites were cut and used to ligate pCIZ into integrating the LTR of the retroviral vector (N2A) (Hantzopoulos et al. 1989). The regions covered by the neomycin and Northern, and Southern probes are shown. Half arrows indicate the location of inverse primers used to find the integration site within the genome of HEK293 cells. U3, R, and U5 make up the LTR of the construct. Expression of the pCIZ trans-splicing molecule is driven by a CMV promoter and includes an intronic region, a BPS, PyT, a consensus 3′ss, and a 3′ exon. The exon here is a partial sequence from LacZ and PPI designates the rat PPI signal.

The trans-splicing construct (pCIZ) was cloned into the LTR (long terminal repeat) region of a Moloney murine leukemia virus (MoMLV) retroviral vector (N2A) (Fig. 1C; Hantzopoulos et al. 1989). Following packaging of N2A-pCIZ into amphotropic viral particles (Laboratory of Dr. Garry Nolan, http://www.stanford.edu/group/nolan/retroviral_systems/retsys.html), human HEK293 cells were infected at a low multiplicity of infection. For our experiments, we used G418 selection to isolate HEK293 cells containing a single integrated N2A-pCIZ. In this article, we investigate the ability of trans-splicing molecules expressed from various integration sites to perform trans-splicing with RNA from genes located near the site of integration.

Proximity can impact trans-splicing to pre-mRNA substrates

A retroviral vector N2A (Hantzopoulos et al. 1989) based on the MoMLV was used to integrate, stably and randomly, an expression cassette for a 3′ PTM (Fig. 1C) into the genome of HEK293 cells. Because the N2A vector contains the neomycin gene, positively infected cells were selected by growth in media containing G418. Through ring cloning, clonal populations were established, and Southern blot analysis was used to identify clonal populations containing a single integrated vector (Fig. 2). Clonal lines containing the vector-derived constant band (4.5 kb) and one other sized band were chosen for further studies (Z2, Z8, Z11, Z13, Z16) (Fig. 2).

FIGURE 2.

(A) Integration of the retroviral vector carrying the trans-splicing molecule. Southern blot showing unique integration bands from several clonal populations. Genomic DNA was digested with BglII and hybridized with a probe covering the intron/LacZ region (Fig. 1c). The 4.5-kb band is a constant band throughout the populations. Populations used for further investigation are in bold and marked with an asterisk. (B) Scheme showing integrating retrovirus with the trans-splicing construct and how the constant and variable bands are derived. The variable band (>1.8 kb) will be 1.8 kb plus the size of the genomic piece cut with BglII. The constant band (4.5 kb) and a probe to the neomycin (Neo) gene (data not shown) were used as Southern controls. The pCIZ integrated construct is represented with a “CMV” rectangle, a black bar (intron region), and a “Z” rectangle (LacZ and the PPI, pre-proinsulin polyadenylation signal region).

After determining which clonal lines had single and unique vector integration sites, we next wanted to determine the vector's exact site of integration in genomic DNA. Using inverse PCR, genomic DNA from each clonal line was digested with BglII and the cleavage products ligated under conditions promoting intramolecular ligation. We performed a nested inverse PCR reaction using primers located within the trans-splicing construct (Fig. 1C). DNA fragments resulting from these reactions were purified and sequenced with both forward and reverse primers. The genomic DNA sequence flanking the vector was then identified using the BLAT search program (UCSC Genome Bioinformatics, http://genome.ucsc.edu/cgi-bin/hgBlat, 05/04 and 06/06 databases). By following the sequence of the vector strand into the genomic sequence, we could determine the orientation of the retrovirus within the genome. Without knowing the location and orientation of the PTM within the genome, some of the spliced products observed could be attributed to cis-splicing of a long pre-mRNA transcript of both the endogenous gene and the PTM sequence (Fig. 3).

FIGURE 3.

Diagram of integration sites of N2A-pCIZ. The population integrated into each gene is noted. The gene name, chromosome position, and total number of exons within the gene are noted for each schematic given. The sizes of exons (squares) and introns (bars) are shown. The integration site and direction are indicated by the small dark gray box (LacZ/PPI) and by the small checkered box with an arrow (CMV). Small half-arrows show the position of primers used in PCR reactions. The checkered/open bars show exons (checkered bar) trans-spliced and tagged with the LacZ exon (open bar tail). The endogenous 5′ss sequence (EXON | intron) involved in trans-slicing are also shown. Bases matching the cis-splicing consensus 5′ donor sequence (^C _AAG | gt^a _gagt) are underlined bold. The largest intron within the gene is indicated with italic numbers.

This scenario was ruled out for populations Z8 and Z16 because the PTM is integrated opposite to the direction of the endogenous gene expression. Therefore, the sequence of the LacZ 3′ss would no longer be recognized as a splice site in a pre-mRNA that included endogenous and PTM sequence. For Z2 the integrant is located within intron 1 and in the same direction as the endogenous gene expression (Fig. 3). If splicing of exon 1 onto LacZ was observed, this could be explained by cis-splicing of a long endogenous/PTM transcript. However, we observed LacZ spliced onto exon 2, which could only be explained by trans-splicing. The scenario for population Z13 is similar to Z2. For Z13, however, we could detect splicing of LacZ onto both exon 1 and exon 2 (Fig. 3). The exon 2/LacZ transcript must come from trans-splicing since the integrant is located within intron 1. The exon 1/LacZ transcript could be either cis- or trans-splicing for population Z13. The exon 1/LacZ transcript observed from the other four populations can only be explained by trans-splicing. Therefore, we surmise the exon 1/LacZ transcript from Z13 is most likely due to trans-splicing or a combination of cis- and trans-splicing. As expected with MoMLV-based gene transfer, all five clonal lines contain vectors stably integrated at different locations in the genome. Moreover, three of these four intragene integration sites are within an intron near the 5′ end of the gene. These results are in agreement with previous studies (Rohdewohld et al. 1987; Scherdin et al. 1990; Wu and Burgess 2004) demonstrating MoMLV prefers to integrate near the 5′ end of active genes.

Next, we wanted to assess the expression levels of the pre-trans-splicing molecule (PTM) in these five clonal lines. First, RNA blot analysis was performed on several populations to visualize the presence of the PTM transcript (data not shown). Second, we used real-time RT-PCR to quantitate PTM RNA expression (Table 1). Total PTM (reacted or unreacted) expression levels were found to be similar within each of the five clonal cell lines tested. With an input of 10 ng mRNA, there was an average of 0.29 pg PTM (unreacted or reacted). For comparison, from this same amount of input, we found an average of 2.3 pg GAPDH (Table 1). Interestingly, when we examined the expression level of the PTM that remained unreacted we observed small differences (Table 1). Clonal populations Z2, Z11, and Z16 had an average of 27% unreacted PTM, while Z8 and Z13 had an average of 40% unreacted PTM.

TABLE 1.

Two-step real-time PCR analysis of pCIZ from cell populations shows similar expression levels for the pre-trans-splicing molecule (PTM)

After determining the genomic location of the integrant and confirming the PTM is similarly expressed, we next wanted to examine if the PTMs were being trans-spliced onto cellular RNAs. Hypothesizing proximity would influence trans-splicing activity, we chose to initially evaluate the ability of PTMs to trans-splice with cellular RNAs expected to be in the immediate vicinity of the PTMs. Since the PTM expression cassettes were integrated into known genes in four of the clonal lines (Fig. 3), we first asked if the PTMs were trans-splicing to the pre-mRNAs expressed from the genes at or near the integration sites. Using mRNA from each of the clonal lines, we performed triplicate RT-PCR reactions using a reverse primer (Fig. 1B) within the PTM and a forward primer (Fig. 3) within the gene-specific transcript.

Table 2 summarizes the RT-PCR results. After sequencing the RT-PCR products with the reverse primer, we eliminated the LacZ portion of the sequence and used the remaining sequence in a BLAT search of the human genome as described above. We identified the genomic sequence and compared it to that in the Ensembl database (www.ensembl.org) to determine the location of the 5′ donor for the trans-spliced event. In all cases, the LacZ exon from the 3′ PTM RNA correctly trans-spliced onto the cellular transcript. Trans-splicing took place at the correct splice site junction of our molecule and involved a natural 5′ donor site from the cellular RNA. In all four cases where the integration occurred within a known gene, trans-splicing to the transcript of this gene could be detected. Trans-splicing to transcripts from the ATP5J, NAV3, and EBF, genes could only be detected in those clonal lines where the integrant was located within that gene (Fig. 4; Table 2). In the other clonal lines, no such trans-splicing product could be detected. In contrast, trans-splicing to ROBO2 RNA could be detected in all five clonal cell lines (Fig. 4; Table 2). Also, in the case of ROBO2, trans-splicing was detected to two different sites within the transcript (Fig. 4; Table 2). When a primer within ROBO2 exon 2 was used in the RT-PCR reactions, only one band was seen for each population. Upon sequencing, this showed trans-splicing had occurred at the exon 2/intron 2 junction. However, when a primer within ROBO2 exon 1 was used in the RT-PCR reactions, only two bands were seen for each population. Sequence analysis showed these bands represented an exon1/LacZ product and an exon 2/LacZ product. The observations above indicate, for trans-splicing to occur with three of the four cellular pre-mRNAs tested, the PTM needed to be expressed in the immediate vicinity of the site of substrate RNA synthesis and processing.

TABLE 2.

RT-PCR results for trans-spliced products from integration site transcripts show some proximity-dependent trans-splicing

FIGURE 4.

(A) Representative agarose gels of trans-spliced RT-PCR products from genes involved in integration. All bands were cut out of gel, cleaned, and sequenced. (Left panel) Small arrowheads indicate trans-spliced products. Other bands are nonspecific RT-PCR products. (Right panel) Diagram shows the location of primers (half arrows) used to find trans-spliced products and a schematic representation of the RT-PCR amplified product. Forward primers to the first exon within each gene and primers to the exon preceding the largest intron were used for the RT-PCR reactions. (B) Example of results obtained from sequencing RT-PCR bands. (Left panel) Sequence chromatogram showing the end of ROBO2 exon2 and the beginning of the LacZ sequence (indicated by an arrow). (Right panel) Example of sequence chromatogram of an unreacted PTM with the intron/LacZ junction indicated by an arrow.

In addition to identifying the location of the trans-spliced events, we also were interested in comparing the amount of cis-spliced and trans-spliced products from transcripts of these PTM-integrated genes. Using biotinylated primers anti-sense to the transcripts from a particular gene and strepavidin-coated beads, we captured both cis- and trans-spliced products from a given amount of mRNA. The captured transcripts were then quantitated by real-time PCR (Table 3). For NAV3 and EBF, there were ∼10 times as many cis-spliced transcripts as trans-spliced transcripts. For ATP5J, this ratio increased to 100-fold while the ratios for ROBO2 ranged from 50 to 200 times more cis-spliced than trans-spliced transcripts.

TABLE 3.

Two-step real-time PCR analysis from cell populations indicates ratio of cis-spliced products to trans-spliced products for each integrant gene transcript

We next wanted to determine whether the pre-mRNAs from genes neighboring the integration site were impacted by proximity to the trans-splicing molecule. Populations Z8 and Z11 both have a PTM construct integrated into chromosome 12. Only the approximate site for the Z11 PTM integration is known, and the distance between this site and the PTM integration gene for Z8 is estimated at 4.4 × 10⁷ bases apart (Fig. 5). As these two integrants were located near each other, we speculated they may trans-splice to the same mRNA substrates. To this end, we looked for genes expressed in HEK293 cells neighboring these PTM loci (Table 4). We tested for trans-spliced products from a gene on either side of NAV3, and another distal to the approximate location of the Z11 PTM (Fig. 5). Though NAV3 was only trans-spliced by the Z8 PTM, the PTMs of both the Z8 and Z11 populations trans-spliced to the transcripts from these three neighbor genes—E2F7, PAWR, and MLXIP.

FIGURE 5.

Genes along chromosome 12. Distance (bp) from Z11 is noted. The asterisks denote trans-spliced transcripts from tables above.

TABLE 4.

RT-PCR results for trans-spliced products from transcripts of genes neighboring integration site show some proximity-dependent trans-splicing

Given these results indicating an influence of proximity on trans-splicing, we examined tagging of genes neighboring the other integrant PTMs. Of these genes, all either were not trans-spliced in any population or were only trans-spliced to the PTM expressed in the immediate vicinity (Table 5). Specifically, clonal line Z2 contains a PTM integrated into the ATP5J gene; trans-splicing to the ATP5J pre-mRNA was only detected in this population. Likewise, trans-splicing to the transcripts from neighboring genes (MRPL39 and GABPA) was also only detected in the Z2 clonal line (Table 5). Similarly, the PTM from clonal line Z16 is integrated into the EBF gene, and trans-splicing to this gene's pre-mRNA was only detected in the Z16 population (Table 5). Trans-splicing to the transcript from ANXA6, the only expressed gene neighboring EBF found to undergo trans-splicing, was also only detected in the Z16 clonal line.

TABLE 5.

RT-PCR results for trans-spliced products from transcripts of genes neighboring integration site also show some proximity-dependent trans-splicing

Colocalization is not essential for spliceosome-mediated RNA trans-splicing to certain substrate pre-mRNAs

In contrast to the proximity-dependent trans-splicing found above, clonal line Z13 contains a PTM integrated into the ROBO2 gene; trans-splicing to the ROBO2 pre-mRNA was detected in all clonal populations (Fig. 4; Table 2). Likewise, trans-splicing to the transcripts from neighboring genes (MITF, ROBO1, and GBE1) was also detected in all cell populations (Table 6). These newly identified targets were good substrates for trans-splicing irrespective of their proximity to the PTM expression site. The differences in trans-spliced products found among the clonal populations was not due to lack of appropriate transcripts. Quantitative RT-PCR showed similar expression levels of each gene among all the clonal populations (Table 7). Again, these observations indicate the importance of proximity of the trans-splicing molecule and certain RNA substrates while trans-splicing to other substrates is proximity independent.

TABLE 6.

RT-PCR results for trans-spliced products also show some proximity-independent trans-splicing

TABLE 7.

Quantitative two-step RT-PCR shows differences in expression levels among genes but similar gene expression levels among clonal populations

DISCUSSION

Since pre-mRNAs are naturally undergoing cis-splicing at their site of transcription, we speculated PTMs would be most effective at trans-splicing to pre-mRNAs if they were expressed near the site of pre-mRNA transcription and processing. Rather than focusing upon a specific pre-mRNA, we took a broad approach to substrate utilization by employing a retroviral vector to randomly integrate a 3′PTM into the human nuclear genome. We then examined trans-splicing to transcripts from the integrant gene as well as to transcripts of genes neighboring the integrant.

Our results demonstrate proximity is important for trans-splicing to occur with a subset of potential substrate pre-mRNAs. The expression of a PTM in close proximity with the ATP5J, NAV3, and EBF, genes was essential for PTM-mediated trans-splicing to occur with the pre-mRNAs expressed from these genes (Fig. 4; Table 2). This trend continued with trans-splicing to neighboring genes (Tables 4, 5). The trans-splicing differences were not due to differences in the number of available PTMs as expression was similar among the populations (Table 1). Interestingly, there were differences in the amount of trans-splicing taking place within the clonal populations. In the Z8 and Z13 populations, the amount of PTM remaining unreacted was ∼40% of the total amount of PTM transcribed. In contrast, populations Z2, Z11, and Z16, had approximately only 27% of the total amount of PTM transcribed remained unreacted. The proximity-dependent trans-splicing is analogous to previous studies where we and others have demonstrated colocalization of catalytic RNAs with their substrate RNAs inside cells enhances the activity of ribozymes (Sullenger and Cech 1993; Lee et al. 1999, 2001). For example, the anti-viral activity of ribozymes is enhanced by appending localization signals to the ribozymes to sort them to the same location as their viral target RNAs (Sullenger and Cech 1993; Lee et al. 1999, 2001). By contrast, in this study, we observed another subset of pre-mRNAs can serve as substrates for trans-splicing regardless of the expression site of the PTM and pre-mRNA transcripts (Tables 4, 5). For each RT-PCR reaction, we used primers from several different exons, but trans-splicing was only observed to the exons described above (Table 2–5). We conclude the majority of transcripts examined were dependent on close proximity to the PTM expression site and within these transcripts, particular exons act as substrates for trans-splicing more readily than others.

Surprisingly, however, we observed PTMs expressed from five different locations could all trans-splice with the pre-mRNA from ROBO2 and its neighboring genes (GBE1, ROBO1, and MITF) even though the five PTM expression sites are located on four different chromosomes. Possible explanations for this result are explained below. HEK293 is an abnormal clonal line, so the ROBO2 chromosome area could be represented a number of times in the genome and thus be more available for trans-splicing than the other gene transcripts. This explanation seems unlikely, however, given the results of later experiments that show all five populations have trans-spliced products from pre-mRNAs from multiple different chromosomes (data not shown). Second, it could be possible all genes on chromosome 3 are proximity independent. The proximity-independent trans-splicing we observed might be a chromosome-specific phenomenon, unrelated to particular genes. This explanation also seems unlikely, as we have observed transcripts from other chromosomes also universally trans-spliced. PLS3 located on the X chromosome has previously been identified as a transcript involved in trans-splicing (Takahara et al. 2005). PLS3/LacZ trans-spliced products were observed in all five of our PTM-integrated populations (data not shown). We have additionally observed universal trans-splicing to transcripts from several other chromosomes: GK also on the X chromosome, NBR2 on chromosome 17, NDUFA4 on chromosome 7, and HSPCB on chromosome 6. Finally, regardless of the location of the integrated PTM (data not shown), the processing site of the ROBO2 area of chromosome 3 may actually be spatially located very near the transcription sites of the PTM genes in the five different clonal lines. Since transcription is largely believed to be coincident with splicing (Maniatis and Tasic 2002; Kornblihtt et al. 2004), the transcription/splicing site of ROBO2, GBE1, ROBO1, and MITF could be near the PTM expression sites in all populations. Thus, the trans-splicing molecules being transcribed from the PTMs integrated within ATP5J, NAV3, and EBF would be near and capable of reacting with the ROBO2 transcripts and the transcripts from its neighboring genes.

Recently, 3D FISH and chromosome conformation capture (3C) were used to assess the organization of several genes along a 40-Mb area of mouse chromosome 7 (Osborne et al. 2004). Here it was found, during active transcription, genes were repositioned into “transcription factories.” Likewise, ROBO2 could be repositioned to an area near the other PTM-integration sites. Recently, a correlation between transcription activity and positioning outside the chromosome territory has been reported (Mahy et al. 2002). The murine Hoxb locus as well as the human IgH and MHC regions typically reside on large loops outside the chromosome territory when poised for transcription (Bickmore and Sumner 1989; Volpi et al. 2000; Ragoczy et al. 2003). These looped chromosome areas are easily reached by the numerous splicing proteins and snRNPs. Similarly, the ROBO2 area may be accessible to the trans-splicing molecules expressed from different chromosomal locations, and not dependent on close proximity to the expression site. These results agree with those of Kikumori et al. (2002), who found trans-splicing between cellular RNAs and adenovirus RNA to be fairly widespread.

In addition to proximity, we speculate some transcripts may be more amenable to trans-splicing regardless of what they are reacting with—adenovirus RNA, LacZ, or another cellular transcript. As shown in Table 8, while looking at the exon/intron makeup of the genes where we demonstrated cellular RNA/LacZ trans-splicing, we noted the following general preferences: (1) The natural 5′ss donor was used in all cases. (2) Trans-splicing took place near the 5′ end of the cellular pre-mRNA; in 47% (n = 8) of those reported here, LacZ was trans-spliced onto the first exon of the cellular transcript. In another 29% (n = 5), the 5′ donor of the second exon was used. The preference for the first two exons holds true even when the PTM and target gene are not linked. The PTM of all populations trans-spliced to the first two exons of ROBO2, though only the PTM from population Z13 was linked to this gene. We have also seen this repeatedly with other genes. For example, transcripts from CBX5 (chromosome 12), NBR2 (chromosome 17), PLS3 and GK (chromosome X), and HSPCB (chromosome 6) are all trans-spliced at the end of exon 1 and/or exon 2 in all of clonal populations (data not shown). (3) The pre-mRNAs generally had extremely large introns, compared with the average human 5-kb intron (Sakharkar et al. 2004). (4) Forty-seven percent (n = 8) of trans-splicing occurred to the exon immediately preceding the largest intron and another 18% (n = 3) involved the second largest intron.

TABLE 8.

Consolidated results for trans-spliced products

These observations are consistent with the findings of other researchers (Kikumori et al. 2002; Takahara et al. 2005). Kikumori et al. (2002) speculated that the size of the intron might delay emergence of the 3′ss long enough for the trans-splicing molecule to invade the spliceosome and compete with cis-splicing. Takahara et al. (2005) examined this idea more closely with the Sp1 transcript. Using Sp1 minigenes, the investigators found trans-splicing depended upon the length of the intron. These investigators also found forcing RNAP II to pause after the 5′ donor site, but before the 3′ acceptor site has been transcribed, increased the likelihood of trans-splicing (Takahara et al. 2005).

In addition to proximity and the preference for large introns, there must be additional factors which promote trans-splicing. For example, the size of the exon replaced may affect trans-splicing. We found in 76.5% (n = 13) of the trans-splicing events, the LacZ of the PTM replaced a smaller-than-average exon (<170 base pairs [bp]) (Sakharkar et al. 2004). Recent statistical analysis finds short exons can be unstable (Weir et al. 2006). With a short exon, the spliceosome complex forming on an upstream 3′ss and downstream 5′ss across an exon may become crowded, sterically hindered, and inefficiently spliced (Dominski and Kole 1991; Weir et al. 2006). This could allow a better opportunity for the trans-splicing molecule to invade. The splicing machinery may find it easier to recognize and use the strong 3′ splicing signals of the PTM than to use the crowded endogenous 3′ss. For gene therapy, these general preferences may be a starting point when choosing an aberrant target RNA with an improved chance at reprogramming with a PTM.

We examined other possible factors that might influence trans-splicing but were unsuccessful at finding additional trends. We speculated a weak 3′ss might promote trans-splicing to the exon preceding this intron. Conversely, a strong 5′ss or a combination of weak 3′ss and strong 5′ss might enhance trans-splicing. However, using MaxEntScan (Laboratory of Dr. Christopher Burge, http://genes.mit.edu/burgelab/maxent/Xmaxentscan_scoreseq.html), we found none of the splice sites in question was particularly strong or weak. There was no detectable pattern with regards to an open reading frame. Takahara et al. (2005) considered the possibility that only trans-spliced products that were in-frame would be detected as these would avoid the nonsense-mediated decay system (Hentze and Kulozik 1999). However, only some of our trans-spliced products preserved the frame or created a large open reading frame. We also considered the possibility the intronic sequence of the trans-splicing molecule may be acting as a binding domain and was preferentially targeting some of the transcripts investigated. In repair PTMs, the binding domain hybridizes to the target sequence, assuring close proximity and increasing the trans-splicing efficiency (Puttaraju et al. 1999, 2001). We found no significant complementarities between our intron and the pre-mRNAs we examined for trans-splicing.

We last considered what affect abundance of transcript may have on whether a transcript was universally trans-spliced or not. Using real-time PCR, we quantitated the expression levels of several genes involved in trans-splicing (Table 7). While the expression level of each transcript was similar among the populations, there were differences in expression levels among the genes (Table 7). Again, however, these did not necessarily provide correlation between trans-splicing differences, though abundance of transcript may be a contributing factor. For example, there is approximately eightfold more ROBO2 transcripts than ATP5J transcripts present within the clonal populations. This may make it easier to trans-splice to ROBO2 transcripts, regardless of PTM integration site. However, there is ∼63-fold more ATP5J transcripts present compared with GBE1, yet trans-splicing to ATP5J is dependent on proximity of the integrated PTM while GBE1 is universally trans-spliced independent of proximity of the integrated PTM. Also, GAPDH, the highest expressing gene we investigated, was not trans-spliced in any of the clonal populations.

The general preferences outlined above and in other studies may be useful when designing repair trans-splicing molecules for gene therapy. Certain pre-mRNAs appear to be better substrates for PTMs than others. Also specific exons within a transcript appear to be more accessible for targeting than others. Finally, more effective PTMs may be ones designed to react with exons near the beginning of the transcript, exons that are followed by a large intron, and PTMs that will replace a small exon.

MATERIALS AND METHODS

HEK293 cell culture

HEK293 cells were cultured at 37°C in growth media, defined as Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin (P-S). Infected HEK293 cells were cultured in DMEM with 10% FBS, 1% P-S, and 1 mg/mL (active units) Geneticin (Invitrogen Corp.).

Trans-splicing construct

The trans-splicing molecule consists of CMV promoter, human adenovirus type 2 intron a partial LacZ sequence as the 3′ exon, and a rat PPI signal (Konarska et al. 1985; Puttaraju et al. 1999). This trans-splicing molecule was digested with BglII and MluI and cloned into the LTR of an N2 retroviral vector. This retroviral vector is a double-copy vector derived from the MoMLV (Hantzopoulos et al. 1989).

Preparing retrovirus

Phoenix cells (http://www.stanford.edu/group/nolan/retroviral_systems/retsys.html) were used to produce amphotropic retroviruses. Phoenix cells were first passaged for 2 wk in hygromycin (300 μg/mL) and diphtheria toxin (1ug/mL) as recommended (http://www.stanford.edu/group/nolan/protocols/pro_helper_dep.html); 1.5 × 10⁶ cells were seeded in six-well tissue culture plates and grown overnight at 37°C in 2 mL growth media (DMEM supplemented with 10% FBS and 1% P-S). Transfection was carried out using 2 μg vector DNA and 10 μL ExGen 500 according to the manufacturer's protocol (Fermentas Inc.). Briefly, DNA/ExGen/NaCl solution was added to growth media followed by a 5-min, 1500-rpm spin and then incubated for 3 h at 37°C. Supernatant was collected 48-h post-transfection and cleaned of cell debris by centrifuging for 5 min at 1500 rpm followed by 0.45-μm syringe filtration. Viral supernatant was used immediately. Transfection efficiencies of >95% were estimated as determined by a FACS analysis of a parallel transfection of 0.2 μg GFP (pEGFP-N1 plasmid from Clontech) and 1.8 μg control vector (pUC19).

Infection

Using DMEM with 10% FBS and 1% P-S, serial dilutions of the retroviral supernatant were made; 1.5 × 10⁶ HEK293 cells were seeded into six-well tissue culture plates and allowed to attach for 4 h. Once attached, fresh media containing the retrovirus or control (no virus) was added to each well along with polybrene (final concentration 4ug/mL). Virus-containing medium was replaced with fresh medium after 5 h of adherence. At 36-h post-infection, medium containing 1mg/mL G418 was added to each well. Approximately 4-wk post-infection, HEK293 clonal populations were created and expanded using the ring cloning technique. Briefly, 5 d after control cells were completely dead, cells were watched closely for growth on G418. Once a single cell had divided and created a visibly distinct colony (∼2 mm in diameter), a glass ring was placed around the colony. This isolated, clonal group of cells was removed and placed into one well of a 24-well tissue culture plate. Clonal lines were expanded and maintained constantly in media containing G418.

RNA and DNA isolation

DNA from each clonal population was isolated using the Puregene DNA isolation kit (Gentra Systems, Inc.). Total RNA was obtained using an RNeasy kit (Qiagen Inc.). mRNA was isolated using the PolyATract mRNA Isolation System (Promega Corp.).

Quantitative two-step RT-PCR

For PTM and gene expression analysis, in a 25 μL reaction, superscript III and a gene-specific primer was added to 50 ng mRNA to create cDNA according to the manufacturer's protocol (Invitrogen Corp.). The reverse transcription was allowed to proceed for 1 h at 55°C; 2× iQ Sybr Green Supermix (Bio-Rad Laboratories) with appropriate primers was added to the 10 μL cDNA reaction to a final 1× Sybr concentration and a final 50 μL volume. For standards, separate PCR reactions were performed followed by gel purification of the amplified band. After quantification on a Nanodrop spectrophotometer (Nanodrop Technologies), serial dilutions of the purified PCR reaction were made and used for standards. Duplicate standard and experimental real-time amplification reactions proceeded through 45 cycles on the Bio-Rad icycler system under the following cycling conditions: 15 sec at 94°C, 20 sec at 57°C, and 40 sec at 72°C. Real-time information was collected during the 72°C extension. A melt curve was generated for each sample reaction after every real-time program.

For comparison of trans-spliced and cis-spliced products (ATP5J, NAV3, EBF, and ROBO2), the Superscript III one-step RT-PCR system (Invitrogen Corp.) was used. Five microliters of captured RNA (ATP5J, NAV3, and EBF, below) or 50 ng mRNA (ROBO2) was used in a 35 μL one-step reaction. RT-PCR proceeded for seven cycles under the following cycling conditions: 15 sec at 94°C, 20 sec at 56°C, and 30 sec at 72°C. After seven cycles, 15 μL of this reaction was used in a quantitative PCR reaction using iQ Sybr Green Supermix (Bio-Rad Laboratories, described above). Standards were created as described above and also proceeded through seven cycles of PCR. For real-time analysis, uncycled standards were used to quantitate cycled standards, and then the cycled standards were used to quantitate experimental products.

Magnetic bead pull-down

Streptavidin-coated Magnaspheres (Promega Corp.) were used to capture both cis- and trans-spliced transcripts for comparison of these products. The manufacturer's protocol was used with a few modifications. Briefly, 1 μg mRNA in a 500 μL volume was heated for 10 min at 65°C; 150 pmol denatured, biotinylated, gene-specific, anti-sense primer was added to the RNA along with 20× SSC to a final concentration of 2×. The primer was allowed to hybridize for 15 min at 50°C. Magnaspheres were captured and washed three times with 0.5 mL 0.5× SSC. Washed beads were incubated with the RNA–primer mixture for 10 min at room temperature. Beads were captured and washed four times with 0.3 mL 0.1× SSC. RNA was eluted with 500 μL nuclease-free water, precipitated overnight, and resuspended in 100 μL nuclease-free water. Five microliters of this RNA was used in quantitative PCR reactions to evaluate cis- and trans-spliced products (see above).

Seminested RT-PCR

Each clonal population was evaluated for the presence of the specific trans-spliced messages. RT-PCR reactions were carried out in triplicate using 40 ng mRNA and the Superscript III one-step RT-PCR system (Invitrogen Corp.). RT-PCR proceeded in a 25 μL reaction for a total of 20 cycles under the following cycling conditions: 15 sec at 94°C, 20 sec at 56°C, and 40 sec at 72°C. After 20 cycles, 1 μL of this reaction was used in a semi-nested PCR reaction. The gene-specific forward primer remained the same while an inner reverse primer was used in a second standard PCR reaction for 25–35 cycles using the same conditions above. Two microliters of this second reaction was electrophoresed on a 2% TBE agarose gel. With each replicate, DNA from each unique PCR band was purified (Qiagen gel extraction kit, Qiagen Inc.) and sequenced to confirm trans-splicing events. The PCR bands were sequenced with either the gene-specific forward primer used for the PCR reaction or a LacZ primer, or both. The resulting sequences contained gene-specific sequence followed by LacZ sequence beginning at the intron/LacZ junction. This “unknown” gene-specific sequence was used in a BLAT search to confirm the gene's identity. The Ensembl database sequence was compared to that of the “unknown” to determine trans-splicing location. For transcripts of genes trans-spliced in multiple different populations, a representative PCR band was purified and sequenced. With each replicate the representative band was isolated from a different population, such that a PCR band from each population was purified and sequenced at least once.

Inverse nested PCR

DNA from each clonal population was digested overnight with BglII, EcoRI, or both restriction enzyme and cleaned using Qiaquick PCR purification kit (Qiagen Inc.); 0.5 μg digested DNA was ligated overnight at 15°C using 2 U ligase in a 500 μL total reaction volume. After ethanol/ammonium acetate precipitation, ligated DNA was resuspended in 100 μL nuclease-free water. Ten microliters was used in a 50 μL PCR reaction using Advantage 2 polymerase (Clontech Laboratories, Inc.). The first PCR was carried out for 25 cycles using the following conditions: 15 sec at 94°C, 20 sec at 57°C, and 7 min at 72°C. Two microliters of this reaction was used in a 50 μL nested PCR reaction. The second PCR proceeded for 35 cycles using the same cycling conditions above. PCR products were electrophoresed; bands were excised and purified from the gel (Qiagen gel extraction kit, Qiagen Inc.) and sequenced with both forward and reverse primers.